NAFEMS Challenge

Andrea

Hi friends! I propose the challenge from NAFEMS. On the WEB is possible to find a solution proposed by a British FEM specialist. I solved the  problem and I achieved a good accordance with their FEA results. After that the solution proposed by the specialist is 231 KN/m2 (to the collapse). I stopped my simulation at about 200 KN/m2 and this is my proposed limit for the plate...maybe I'm too on safe!!!
Try to solve it.....

Andrea

P.S. I done it with mecway. If anyone needs LIML file I will post it

Victor

Just curious what makes this problem challenging. Is it hard for solvers to converge? Or something complicated about the boundary conditions or the load changes while it's deforming?

Andrea

Conceptually is not complicated but is a NAFEMS challenge. So is necessary to perform a FEA considering all the steps. For example:
a) Boundary: perform linear static to know if boundary condition will affect the results. If boundary are iperstatic could affect the results in a different manner performing second order analysis. Sensitivity studies must be performed to find the better boundary
b) calibration against known values, when is possible
c) mesh convergence studies for stress analysis
d) time/load increment sufficient small (convergence study)
e) check the reaction
f) hourglass control for reduced integration (if used)
.......and other

In this case if one performs a static analysis locking vertical displacements of perimetral nodes will find a downward vertical reaction at the corner  which is not real. How this affects the results?
If one use CCX with shell elements S8R must evaluate hourglass energy (must remains small)
If one use CCX S8 elements must know that this elements are expanded to a 20 nodes brick elements. So if one restraints a node of the shell this node is the middle node through thickness. Conceptually  must be restrained  the lower nodes. How this affects the results? For example is necessary to conduct a study adding OFFSET parameter to *SHELL SECTION
Gravity load: is it negligible?

Stefanos

Andrea, could you please post the LIML file

Andrea

I attach file relative to static sensitivity analisys and mesh convergence. You will find 3 plates: first and second the real plate with different boundary. The third a plate with a/b = 3. This ratio, not very different from 2000/600 of the real plates, allows rapid comparison with theory results (see Roark and Young text: Formulas for stress and strain). Theory results are: max s = 317 MPa, max z = -12,56 mm. So the mesh is sufficient fine. After that is possible to be sure that different boundary doesn't affect the results. In fact the first and the second plates shows very closed value of s and displacement. (note the different node reaction at the corner nodes). Load is 100 kPa uniform pressure. So with this mesh and boundary (see plate 1) is possible to continue the study first non linear geometry and after with nonlinear material (bilinear s-e true stress/true strain curve)